Apple’s presentation of the M2 SoC was done with great fanfare, but the truth is that hiding the obvious is making many realize that marketing can’t hide shame. And that is precisely what happens when you analyze this SoC and the feeling that the Apple M2 and its architecture: disappointment. Why did Apple launch such a chip knowing that the competition is getting tougher?
The reasons are basically a flight of engineers in search of more interesting projects and above all money. It is assumed that Apple does not pay its employees very well, while the competition does, where it also seems that the structural organization of Google, Amazon or Nuvia, for example, is better. This and two years of work have caused an easily understandable debacle considering what has been said.
Apple M2, architecture: the disappointment that should never see the light of day
We assume we have the cool, fresh SoC features in mind, so let’s jump right into the architecture. As for the size of the dice, we are talking about 155.25mm2 according to estimates that have been made from Locuza and Semianalysis, where in this case and with the 20 billion transistors the taking into account of the density would have, curiously, decreased. Something surprising given that TSMC’s N5+ was denser.
Focusing on the P-Core, we see an impressive 21% increase in size over the M1. This is favored by the larger size of the L2 when growing at 16 MB, which also requires more space for the internal buses due to the higher bandwidth. Apple is widely believed to have spent 5.2 mm2 in the P-Corewhere curiously and since there are no significant changes beyond the L2, the increase in the CPI comes from the frequency, which is really disappointing.
Well, the disappointment does not end there, since the E-Cores are the same as in the M1, there is no change, or at least it shows under the microscope, since there is barely 1 mm2 difference.
The graphics card, a facelift to affirm that it is made for gaming
The biggest change comes from the GPU and with it Apple’s biggest defeat in this M2. increases to 10 hearts counting, increases the frequency by 1.27 GHz to 1.406GHz and the size of it in 7 mm2, but there are no more changes. So it’s more or less the same with two cores added and more space consumed, with higher consumption and frequency, nothing more. The performance comes from here and not as Apple wanted to sell a new GPU, because neither in the architecture nor in the buses there is a difference.
The last important aspect to analyze is the RAM memory. The total area of the chip concentrated on this section with 128 bits are 14 mm2 compared to the 8.1 of the original M1 and it is not that it is important in itself, but rather that this increase comes from the use LPDDR5 6400 compared to LPDDR4X at 4266 MHz. The problem is that it’s the same bus for higher bandwidth with higher RAM speed, so the power cost goes up, like the final price of this M2 SoC, because LPDDR5 is much more expensive, which in turn will make iPhones and Macs more expensive of course.
As we can see, the changes are minimal, it is a facelift that will cost more, consume more and in terms of performance it is by no means a panacea. Two years of work is a long time to present this Apple M2 with this architecture as such. We all expected a lot more, like including architecture ARMv9or to reach 3nm with TSMC as happened with the M1 and its 5 nm. It’s going to cost Apple more than one review and we’ll have to wait for the M3 to see if the changes are worth it to justify the prices that the bitten apple ones will introduce to the market.